Ramp short-term traffic flow prediction based on BiLSTM-Attention model with spatiotemporal correlation analysis

Konferenz: ISCTT 2021 - 6th International Conference on Information Science, Computer Technology and Transportation
26.11.2021 - 28.11.2021 in Xishuangbanna, China

Tagungsband: ISCTT 2021

Seiten: 6Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Gong, Yannan; Lu, Wenqi; Rui, Yikang; Ran, Bin (School of Transportation, Southeast University, Nanjing, China & Jiangsu Key Laboratory of Urban ITS, Southeast University, Nanjing, China & Joint Research Institute on Internet of Mobility, Southeast University and University of Wisconsin-Madison, Nanjing, China)

On-off ramp control of expressway is a key component in the field of traffic management and control, which is of great significance for improving the service level of expressway. In order to improve the fine management ability of expressway, and consider the spatio-temporal influence relationship between expressway main line and ramp as a whole, this paper introduced BiLSTM-Attention model on the basis of mining the correlation between the traffic flow of off-ramp and upstream interweaving area. This method first adopted the random forest algorithm to obtain the relevant lane which have the greatest correlation with the predicted ramp, and used it as the model inputs, and then the BiLSTM-Attention model was constructed for short-term traffic flow prediction. Compared with the traditional model, this model took into account the spatio-temporal influence of interweaving area of the multi-lane expressway on off-ramp. At the same time, BiLSTM model was used to obtain the spatio-temporal characteristics of upstream and downstream traffic flow, and Attention mechanism was introduced to learn the key information of time series, which effectively improves the prediction accuracy of short-term traffic flow. This method was used to predict the traffic flow of an off-ramp section of the California Expressway. The results show that the root mean square error (RMSE) and mean absolute error (MAE) of this method are less than the comparison model, therefore, this method has better effect.